
# System
import warnings
warnings.filterwarnings("ignore")

# Data manipulation
import pandas as pd
pd.get_option("display.max_columns")
import numpy as np

# Visualizations
import matplotlib.pyplot as plt
import seaborn as sns

# Statistics
import statsmodels.api as sm
from statsmodels.stats.anova import AnovaRM
from pingouin import ttest
from scipy import stats
from scipy.stats import median_test
from sklearn.linear_model import LinearRegression


# The following three functions make a correlation matrix with scatterplot in lower left, 
# pearson correlation coefficient bubble plot in upper right, and distribution
# plot down the diagonal
def corrdot(*args, **kwargs):
    corr_r = args[0].corr(args[1], 'pearson')
    corr_text = f"{corr_r:2.2f}".replace("0.", ".")
    ax = plt.gca()
    ax.set_axis_off()
    marker_size = 1
    ax.scatter([.5], [.5], marker_size, [corr_r], alpha=0.001, cmap="Greys", \
               vmin=-1, vmax=1, transform=ax.transAxes)
    font_size = 40
    ax.annotate(corr_text, [.5, .5,],  xycoords="axes fraction", ha='center', \
                va='center', fontsize=font_size)

def annotate_colname(x, **kws):
    ax = plt.gca()
    ax.annotate(x.name, xy=(0.05, 0.9), xycoords=ax.transAxes, fontweight='bold')
    
def corr_plot(df):
    sns.set(style='white', font_scale=1.6)
    g = sns.PairGrid(df, aspect=1.4, diag_sharey=False)
    g.map_lower(sns.regplot, scatter_kws={'s':10, 
                                          'color':'k'})
    g.map_diag(sns.distplot, kde_kws={'color': 'black'})
    g.map_diag(annotate_colname)
    g.map_upper(corrdot)
    plt.rcParams["figure.dpi"] = 300
    plt.show()

def theory_cat(df, col):
    cat = []
    for i in range(len(df)):
        if df[col][i]>2:
            categ = 'low'
        else:
            categ = 'high'
        cat.append(str(col+"_"+categ))
    return cat


raw_data = pd.read_csv("../data/01_raw/raw_for_analysis.csv")
data = raw_data.copy()


data.sample(5)


for col in ['race', 'education', 'income']:
    temp_df = data[col].value_counts().reset_index()
    temp_df['percentage'] = (temp_df[col]/temp_df[col].sum())*100
    print(col, temp_df, "\n\n")

race                                               index  race  percentage
0                                White or Caucasian    25   83.333333
1                                             Asian     1    3.333333
2                          White or Caucasian,Other     1    3.333333
3         Native Hawaiian or Other Pacific Islander     1    3.333333
4                         Black or African American     1    3.333333
5  American Indian/Native American or Alaska Native     1    3.333333 


education                                                index  education  percentage
0                                  Bachelor's degree         14   46.666667
1                        Some college, but no degree          7   23.333333
2  Graduate or professional degree (MA, MS, MBA, ...          6   20.000000
3                         High school diploma or GED          3   10.000000 


income                index  income  percentage
0    $25,000-$49,999       7   23.333333
1    $75,000-$99,999       7   23.333333
2  $100,000-$149,999       5   16.666667
3  Less than $25,000       5   16.666667
4    $50,000-$74,999       4   13.333333
5  Prefer not to say       1    3.333333
6   $150,000 or more       1    3.333333


# Subset the dataframe to only the cols with the primary DVs. 
df = data[[
    'race',
    'education',
    'income',
    'one_ten_prop_10',
    'five_one_prop_50',
    'one_five_prop_10',
    'ten_one_prop_100',
    'loss_money_h',
    'gain_money_h',
    'loss_ethics_h',
    'gain_ethics_h',
    'utilitarian_rank',
    'deontology_rank',
    'virtue_rank',
    'ethics_care_rank',
    'contract_theory_rank'
]]

# Check out a snippet of the resulting df
df.sample(5)


# Plot all of it
corr_plot(df)


prob_match_df = df[['one_ten_prop_10', 'one_five_prop_10', 'five_one_prop_50', 'ten_one_prop_100']]


# Correlation matrix
corr_plot(prob_match_df)


# PREP FOR EASIER PLOTTING
prob_match_df_long = prob_match_df.copy().dropna().reset_index(drop=True)
prob_match_df_long['p_id'] = list(range(1, len(prob_match_df)+1))

# Turn from wide to long
prob_match_df_long = prob_match_df_long.melt(id_vars=['p_id'])

# Replace header vals with expected value of first option
prob_match_df_long = prob_match_df_long.replace(           
    {
        'one_ten_prop_10':.9, 
        'one_five_prop_10':.8,
        'five_one_prop_50':.2, 
        'ten_one_prop_100':.1
    }
)
prob_match_df_long[::17]


# Scatterplot
fig, ax = plt.subplots(figsize=(10, 10))
sns.regplot(x='variable', y='value', data=prob_match_df_long, 
            scatter_kws={"color": "k"}, line_kws={"color": "black"})
plt.ylabel("Proportion First Option Chosen", fontsize=30, labelpad=20)
plt.xlabel("Expected Value of First Option", fontsize=30, labelpad=20)
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
plt.show()


# Boxplot
fig, ax = plt.subplots(figsize=(10, 10))
sns.swarmplot(x='variable', y='value', data=prob_match_df_long, color='k')
sns.boxplot(x='variable', y='value', data=prob_match_df_long, color='w')
plt.ylabel("Proportion First Option Chosen", fontsize=30, labelpad=12)
plt.xlabel("Expected Value of First Option", fontsize=30, labelpad=12)
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
plt.rcParams["figure.dpi"] = 300
plt.show()


# Repeated measures ANOVA
model = AnovaRM(data=prob_match_df_long, depvar='value', subject='p_id', within=['variable'])
res = model.fit()
res.summary()


# Save the results
ttest_prob_match = pd.DataFrame()

# Paired t-tests
for col_1 in list(prob_match_df):
    for col_2 in list(prob_match_df):
        if col_1 == col_2:
            continue
        else:
            res = ttest(prob_match_df[col_1], 
                        prob_match_df[col_2], 
                        paired=True, 
                        alternative='two-sided')
            res['first_col'] = col_1
            res['second_col'] = col_2
            ttest_prob_match = pd.concat([ttest_prob_match, res])
            
ttest_prob_match = ttest_prob_match.reset_index(drop=True)
ttest_prob_match = ttest_prob_match[[ 'first_col', 'second_col', 'T', 'dof',
                                     'alternative', 'p-val', 'CI95%', 'cohen-d',
                                     'BF10', 'power']]
ttest_prob_match['p-val'] = round(ttest_prob_match['p-val'], 6)
ttest_prob_match


# Scatterplot of just average with errorbars
y_avg = prob_match_df_long.groupby(by=['variable']).mean().reset_index()['value']
err = prob_match_df_long.groupby(by=['variable']).sem().reset_index()['value']

fig, ax = plt.subplots(figsize=(10, 10))
sns.regplot(x='variable', y='value', data=prob_match_df_long, 
            scatter_kws={"color": "w"}, line_kws={"color": "black"})
plt.errorbar(x=[.1, .2, .8, .9], y=y_avg, yerr=err, fmt='o', color='k')
sns.lineplot(x=[0, 1], y=[0, 1], color='k', alpha=0.5, linestyle='dashed')
plt.ylabel(r"$\dfrac{B_{1}}{B_{1}+B_{2}}$", fontsize=30, labelpad=20)
ticks = [0.0, .2, .4, .6, .8, 1.0]
plt.yticks(ticks=ticks, labels=ticks, fontsize=20)
plt.xticks(ticks=ticks, labels=ticks, fontsize=20)
plt.xlim(-0.075, 1.075)
plt.xlabel(r'$\dfrac{r_{1}}{r_{1}+r_{2}}$', fontsize=30, labelpad=20)
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
plt.rcParams["figure.dpi"] = 300
plt.show()


x = list(np.log([0.1, 0.2, 0.8, 0.9]).reshape(-1, 1))
scores = []
slopes = []
y_int = []

for i in range(len(prob_match_df)):
    y = list(prob_match_df.iloc[i])
    y = [val+0.1 for val in y]
    y = np.log(y)
    try:
        reg = LinearRegression().fit(x,y)
        scores.append(reg.score(x, y))
        slopes.append(np.abs(reg.coef_[0]))
        y_int.append(reg.intercept_)
    except:
        continue

fit_data = {'gme_r2':scores, 
            'gme_slopes':slopes, 
            'gme_y_int':y_int}

fit_data = pd.DataFrame(fit_data)
fit_data[::7]


# Boxplot
fig, ax = plt.subplots(nrows=3, ncols=1, figsize=(3, 15))
# r2
sns.swarmplot(ax=ax[0], y=fit_data['gme_r2'], color='k', size=6)
sns.boxplot(ax=ax[0], y=fit_data['gme_r2'], color='w')
ax[0].set_ylim(0, 1)
ax[0].set_ylabel("$r^2$", fontsize=40)
ax[0].yaxis.set_tick_params(labelsize=20)
ax[0].spines.right.set_visible(False)
ax[0].spines.top.set_visible(False)

# Slopes
sns.swarmplot(ax=ax[1], y=fit_data['gme_slopes'], color='k', size=6)
sns.boxplot(ax=ax[1], y=fit_data['gme_slopes'], color='w')
ax[1].set_ylim(0, 1.35)
ax[1].set_ylabel("Sensitivity", fontsize=40)
ax[1].yaxis.set_tick_params(labelsize=20)
ax[1].spines.right.set_visible(False)
ax[1].spines.top.set_visible(False)

# Y-int
sns.swarmplot(ax=ax[2], y=fit_data['gme_y_int'], color='k', size=6)
sns.boxplot(ax=ax[2], y=fit_data['gme_y_int'], color='w')
ax[2].set_ylim(-2.8, 0.55)
ax[2].set_ylabel("Bias", fontsize=40)
ax[2].yaxis.set_tick_params(labelsize=20)
ax[2].spines.right.set_visible(False)
ax[2].spines.top.set_visible(False)
plt.rcParams["figure.dpi"] = 300
plt.show()


# What is the proportion where VAC is greater than 0.60?
len(fit_data[fit_data['gme_r2']>=.6])/len(fit_data)

0.6666666666666666


# Append to df
df = pd.concat([df, fit_data], axis=1)
df[::9]


disc_df = df[['loss_money_h', 'gain_money_h', 'loss_ethics_h', 'gain_ethics_h',]]
disc_df.columns = ['Lose Money', 'Gain Money', 'Violate Ethics', 'Comply Ethics']
disc_df = np.log(disc_df)
disc_df.sample(5)


# Correlation matrix
corr_plot(disc_df)


# PREP FOR EASIER PLOTTING
disc_df_long = disc_df.copy()
disc_df_long['p_id'] = list(range(1, len(disc_df_long)+1))

# Turn from wide to long
disc_df_long = disc_df_long.melt(id_vars=['p_id'])


# Repeated measures ANOVA
model = AnovaRM(data=disc_df_long, depvar='value', subject='p_id', within=['variable'])
res = model.fit()
res.summary()


# Save the results
ttest_disc_df = pd.DataFrame()

# Paired t-tests
for col_1 in list(disc_df):
    for col_2 in list(disc_df):
        if col_1 == col_2:
            continue
        else:
            res = ttest(disc_df[col_1], 
                        disc_df[col_2], 
                        paired=True, 
                        alternative='two-sided')
            res['first_col'] = col_1
            res['second_col'] = col_2
            ttest_disc_df = pd.concat([ttest_disc_df, res])
            
ttest_disc_df = ttest_disc_df.reset_index(drop=True)
ttest_disc_df = ttest_disc_df[[ 'first_col', 'second_col', 'T', 'dof',
                                     'alternative', 'p-val', 'CI95%', 'cohen-d',
                                     'BF10', 'power']]
ttest_disc_df


# Boxplot
fig, ax = plt.subplots(figsize=(10, 10))
sns.swarmplot(x='variable', y='value', data=disc_df_long, color='k')
sns.boxplot(x='variable', y='value', data=disc_df_long, color='w', 
            showmeans=True, meanprops={"marker":"*",
                                       "markerfacecolor":"w", 
                                       "markeredgecolor":"black",
                                       "markersize":"20"})
plt.ylabel("$\itlog({h})$", fontsize=30, labelpad=20)
ticks = [0, 1, 2, 3]
labels=['Money\nLoss', 'Money\nGain', 'Ethics\nLoss', 'Ethics\nGain']
plt.xticks(ticks=ticks, labels=labels)
plt.xlabel("Probability Discounting Task", fontsize=30, labelpad=20)
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
plt.rcParams["figure.dpi"] = 300
plt.show()


theor_df = df[['utilitarian_rank', 'deontology_rank', 'virtue_rank', 'ethics_care_rank', 'contract_theory_rank']]
theor_df[::8]


# Correlation matrix
corr_plot(theor_df)


# PREP FOR EASIER PLOTTING
theor_df_long = theor_df.copy()
theor_df_long['p_id'] = list(range(1, len(theor_df)+1))

# Turn from wide to long
theor_df_long = theor_df_long.melt(id_vars=['p_id'])
theor_df_long[::34]


# Repeated measures ANOVA
theor_df_long = theor_df_long.dropna().reset_index(drop=True)
model = AnovaRM(data=theor_df_long, depvar='value', subject='p_id', within=['variable'])
res = model.fit()
res.summary()


# Save the results
ttest_theor_df = pd.DataFrame()

# Paired t-tests
for col_1 in list(theor_df):
    for col_2 in list(theor_df):
        if col_1 == col_2:
            continue
        else:
            res = ttest(theor_df[col_1], 
                        theor_df[col_2], 
                        paired=True, 
                        alternative='two-sided')
            res['first_col'] = col_1
            res['second_col'] = col_2
            ttest_theor_df = pd.concat([ttest_theor_df, res])
            
ttest_theor_df = ttest_theor_df.reset_index(drop=True)
ttest_theor_df = ttest_theor_df[[ 'first_col', 'second_col', 'T', 'dof',
                                     'alternative', 'p-val', 'CI95%', 'cohen-d',
                                     'BF10', 'power']]
ttest_theor_df


theor_df_long


# Boxplot
fig, ax = plt.subplots(figsize=(18, 10))
sns.swarmplot(x='variable', y='value', data=theor_df_long, color='k')
sns.boxplot(x='variable', y='value', data=theor_df_long[theor_df_long['value']<6], color='w', 
           showmeans=True, meanprops={"marker":"*",
                                      "markerfacecolor":"white", 
                                      "markeredgecolor":"black",
                                      "markersize":"20"})
plt.ylabel("Ranking", fontsize=30, labelpad=20)
ticks = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
plt.yticks(ticks=ticks, labels=ticks)
plt.ylim(0.8, 6.2)
ticks = [0, 1, 2, 3, 4]
labels=["Utilitarian", 'Deontology', 'Virtue\nTheory', 'Ethics\nof Care', 'Contract\nTheory']
plt.xticks(ticks=ticks, labels=labels, fontsize=20)
plt.xlabel("Ethical Theory", fontsize=30, labelpad=20)
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
plt.rcParams["figure.dpi"] = 300
plt.show()


theor_df.describe()


# Convert raw rankings to binary category
df['util_rank_cat'] = theory_cat(df=df, col='utilitarian_rank')
df['deont_rank_cat'] = theory_cat(df=df, col='deontology_rank')
df['virtue_rank_cat'] = theory_cat(df=df, col='virtue_rank')
df['care_rank_cat'] = theory_cat(df=df, col='ethics_care_rank')
df['contract_rank_cat'] = theory_cat(df=df, col='contract_theory_rank')


# Arrange for easy plotting
plot_df = df.drop(['race', 'education', 'income', 'utilitarian_rank', 'deontology_rank', 
                   'virtue_rank', 'ethics_care_rank', 'contract_theory_rank', 
                   'one_ten_prop_10', 'one_five_prop_10',
                   'five_one_prop_50', 'ten_one_prop_100'], axis=1)

# Drop those for whom the GME firt was below 0.60
plot_df = plot_df[plot_df['gme_r2']>=0.60].reset_index(drop=True)

plot_df = plot_df.melt(id_vars=['util_rank_cat', 'deont_rank_cat', 
                                'virtue_rank_cat', 'care_rank_cat', 
                                'contract_rank_cat']).reset_index(drop=True)
plot_df


plot_df.variable.unique()

array(['loss_money_h', 'gain_money_h', 'loss_ethics_h', 'gain_ethics_h',
       'gme_r2', 'gme_slopes', 'gme_y_int'], dtype=object)


plot_df_prop = plot_df[plot_df['variable'].isin(['gme_r2', 'gme_slopes','gme_y_int'])]

# Boxplots
theories = ['util_rank_cat', 'deont_rank_cat', 'virtue_rank_cat', 'care_rank_cat', 'contract_rank_cat']

for theor in theories:
    plot_df_prop = plot_df_prop.sort_values(by=['variable', theor])
    fig, ax = plt.subplots(figsize=(10, 10))
    sns.swarmplot(x='variable', y='value', hue=theor, dodge=True, data=plot_df_prop, 
                  palette='binary', order=['gme_r2', 'gme_slopes','gme_y_int'])
    sns.boxplot(x='variable', y='value', hue=theor, data=plot_df_prop, color='w')
    plt.ylabel("Metric Value", fontsize=30, labelpad=20)
    ticks = [0, 1, 2]
    labels=['$r^{2}$', 'Sensitivity', 'Bias']
    plt.xticks(ticks=ticks, labels=labels)
    plt.xlabel("GME Parameter", fontsize=30, labelpad=20)
    ax.spines['top'].set_visible(False)
    ax.spines['right'].set_visible(False)
    plt.legend(loc=(1.05,0.6))
    plt.rcParams["figure.dpi"] = 300
    plt.show()


plot_df_prop


plot_df_prop = plot_df[plot_df['variable'].isin(['loss_money_h', 'gain_money_h', 
                                                 'loss_ethics_h', 'gain_ethics_h'])]
plot_df_prop['value'] = np.log(plot_df_prop['value'])

# Boxplots
theories = ['util_rank_cat', 'deont_rank_cat', 'virtue_rank_cat', 'care_rank_cat', 'contract_rank_cat']

for theor in theories:
    plot_df_prop = plot_df_prop.sort_values(by=['variable', theor])
    fig, ax = plt.subplots(figsize=(10, 10))
    sns.swarmplot(x='variable', y='value', hue=theor, dodge=True, data=plot_df_prop, palette='binary')
    sns.boxplot(x='variable', y='value', hue=theor, data=plot_df_prop, color='w')
    plt.ylabel("$\itlog({h})$", fontsize=30, labelpad=20)
    ticks = [0, 1, 2, 3]
    labels=['Ethics\nGain', 'Money\nGain', 'Ethics\nLoss', 'Money\nLoss']
    plt.xticks(ticks=ticks, labels=labels)
    plt.xlabel("Discounting Task", fontsize=30, labelpad=20)
    ax.spines['top'].set_visible(False)
    ax.spines['right'].set_visible(False)
    plt.legend(loc=(1.05,0.6))
    plt.rcParams["figure.dpi"] = 300
    plt.show()


quant_df = df[['loss_money_h', 'gain_money_h', 'loss_ethics_h', 'gain_ethics_h']]
quant_df = np.log(quant_df)
gme_df = df[['gme_r2', 'gme_slopes', 'gme_y_int']]
quant_df = pd.concat([gme_df, quant_df], axis=1)
quant_df = quant_df[quant_df['gme_r2']>=.8].reset_index(drop=True)
quant_df.columns = ['Prob Match\n$r^2$', 'Prob Match\nSensitivity', 'Prob Match\nBias', 
                    'Lose Money', 'Gain Money', 'Violate Ethics', 'Comply Ethics']
quant_df[::9]


corr_plot(quant_df)


theor_df = df[['loss_ethics_h', 'gain_ethics_h']]
theor_df = np.log(theor_df)
rank_df = df[['utilitarian_rank', 'deontology_rank', 'virtue_rank', 
              'ethics_care_rank', 'contract_theory_rank']]
theor_df = pd.concat([rank_df, theor_df], axis=1)
theor_df.columns = ['Utilitarianism', 'Deontology', 'Virtue Theory', 
                    'Ethics of Care', 'Contract Theory', 
                    'Violate Ethics', 'Comply Ethics']
theor_df = theor_df.dropna().reset_index(drop=True)
theor_df[::9]


theor_df.describe()


corr_plot(theor_df)

	F Value	Num DF	Den DF	Pr > F
variable	21.6534	3.0000	87.0000	0.0000

	first_col	second_col	T	dof	alternative	p-val	CI95%	cohen-d	BF10	power
0	one_ten_prop_10	one_five_prop_10	1.654212	29	two-sided	0.108870	[-0.03, 0.25]	0.356338	0.655	0.470972
1	one_ten_prop_10	five_one_prop_50	5.846852	29	two-sided	0.000002	[0.35, 0.72]	1.748224	7770.783	1.000000
2	one_ten_prop_10	ten_one_prop_100	5.716404	29	two-sided	0.000003	[0.37, 0.78]	1.912417	5576.794	1.000000
3	one_five_prop_10	one_ten_prop_10	-1.654212	29	two-sided	0.108870	[-0.25, 0.03]	0.356338	0.655	0.470972
4	one_five_prop_10	five_one_prop_50	3.913957	29	two-sided	0.000505	[0.2, 0.65]	1.370838	60.705	1.000000
5	one_five_prop_10	ten_one_prop_100	5.635086	29	two-sided	0.000004	[0.3, 0.64]	1.525648	4534.202	1.000000
6	five_one_prop_50	one_ten_prop_10	-5.846852	29	two-sided	0.000002	[-0.72, -0.35]	1.748224	7770.783	1.000000
7	five_one_prop_50	one_five_prop_10	-3.913957	29	two-sided	0.000505	[-0.65, -0.2]	1.370838	60.705	1.000000
8	five_one_prop_50	ten_one_prop_100	0.516794	29	two-sided	0.609219	[-0.12, 0.2]	0.131515	0.22	0.107199
9	ten_one_prop_100	one_ten_prop_10	-5.716404	29	two-sided	0.000003	[-0.78, -0.37]	1.912417	5576.794	1.000000
10	ten_one_prop_100	one_five_prop_10	-5.635086	29	two-sided	0.000004	[-0.64, -0.3]	1.525648	4534.202	1.000000
11	ten_one_prop_100	five_one_prop_50	-0.516794	29	two-sided	0.609219	[-0.2, 0.12]	0.131515	0.22	0.107199

	gme_r2	gme_slopes	gme_y_int
0	0.928515	1.242624	-2.420982
7	0.762375	0.657770	-1.304928
14	0.827343	0.851415	-1.928710
21	0.928515	1.242624	-2.420982
28	0.114643	0.130548	-0.593438

	Lose Money	Gain Money	Violate Ethics	Comply Ethics
29	0.198851	0.470004	-1.469676	0.756122
5	1.238374	0.058269	-3.912023	2.810005
2	0.198851	0.756122	0.058269	-0.462035
19	0.598837	1.456287	-0.061875	0.198851
14	0.756122	0.598837	0.756122	0.198851

	F Value	Num DF	Den DF	Pr > F
variable	7.9245	3.0000	87.0000	0.0001

Packages¶

Functions¶

Read in the data¶

Demographics information¶

Analytic Dataframe¶

Visualizations¶

Probability Matching¶

# Proper probablity matching plot¶

Individual Level¶

Probability Discounting Tasks¶

Ethics Theory Ranking¶

Relationships between different decision-making tasks¶

Differences in matching¶

	race	education	income	one_ten_prop_10	one_ten_q1	one_ten_q2	one_ten_q3	one_ten_q4	one_ten_q5	one_ten_q6	...	other_rank_1	other_rank_1_text	other_rank_2	other_rank_2_text	other_rank_3	other_rank_3_text	other_rank_4	other_rank_4_text	other_rank_5	other_rank_5_text
2	White or Caucasian	Some college, but no degree	$50,000-$74,999	0.7	10	10	10	10	10	0.0	...	3	personal morals	10	environmental impacts	9	build up to the decision	8	number of people impacted	7	background of those involved
10	White or Caucasian	Bachelor's degree	$100,000-$149,999	1.0	10	10	10	10	10	10.0	...	6	NaN	7	NaN	8	NaN	9	NaN	10	NaN
15	White or Caucasian	Graduate or professional degree (MA, MS, MBA, ...	$75,000-$99,999	1.0	10	10	10	10	10	10.0	...	6	NaN	7	NaN	8	NaN	9	NaN	10	NaN
1	White or Caucasian	Bachelor's degree	$25,000-$49,999	0.9	10	10	10	10	10	10.0	...	5	The reasonableness of the action	7	NaN	8	NaN	9	NaN	10	NaN
27	White or Caucasian	High school diploma or GED	$50,000-$74,999	1.0	10	10	10	10	10	10.0	...	6	NaN	7	NaN	8	NaN	9	NaN	10	NaN

	race	education	income	one_ten_prop_10	five_one_prop_50	one_five_prop_10	ten_one_prop_100	loss_money_h	gain_money_h	loss_ethics_h	gain_ethics_h	utilitarian_rank	deontology_rank	virtue_rank	ethics_care_rank	contract_theory_rank
22	Native Hawaiian or Other Pacific Islander	High school diploma or GED	Less than $25,000	0.7	0.3	0.8	0.3	1.06	0.82	0.01	2.93	4	5	1	2	3
1	White or Caucasian	Bachelor's degree	$25,000-$49,999	0.9	0.1	0.9	0.1	0.94	1.22	1.06	0.82	1	3	2	6	4
11	White or Caucasian	Graduate or professional degree (MA, MS, MBA, ...	$25,000-$49,999	0.1	0.8	0.1	0.9	0.55	2.13	0.01	1.82	4	3	2	6	5
7	White or Caucasian	Graduate or professional degree (MA, MS, MBA, ...	$75,000-$99,999	0.9	0.1	1.0	0.3	4.29	2.13	1.22	3.45	4	1	2	5	3
27	White or Caucasian	High school diploma or GED	$50,000-$74,999	1.0	0.0	1.0	0.0	3.45	9.65	3.45	0.18	3	1	2	5	4

	p_id	variable	value
0	1	0.9	1.0
17	18	0.9	1.0
34	5	0.8	0.4
51	22	0.8	1.0
68	9	0.2	0.3
85	26	0.2	0.2
102	13	0.1	0.0
119	30	0.1	0.0

	first_col	second_col	T	dof	alternative	p-val	CI95%	cohen-d	BF10	power
0	Lose Money	Gain Money	-0.912361	29	two-sided	0.369101	[-0.79, 0.3]	0.212714	0.285	0.203373
1	Lose Money	Violate Ethics	3.606569	29	two-sided	0.001151	[0.64, 2.31]	0.870246	29.275	0.995899
2	Lose Money	Comply Ethics	0.091970	29	two-sided	0.927354	[-0.64, 0.7]	0.024552	0.195	0.051941
3	Gain Money	Lose Money	0.912361	29	two-sided	0.369101	[-0.3, 0.79]	0.212714	0.285	0.203373
4	Gain Money	Violate Ethics	4.629200	29	two-sided	0.000071	[0.96, 2.48]	0.955345	353.708	0.999009
5	Gain Money	Comply Ethics	0.665850	29	two-sided	0.510769	[-0.57, 1.11]	0.199690	0.238	0.184754
6	Violate Ethics	Lose Money	-3.606569	29	two-sided	0.001151	[-2.31, -0.64]	0.870246	29.275	0.995899
7	Violate Ethics	Gain Money	-4.629200	29	two-sided	0.000071	[-2.48, -0.96]	0.955345	353.708	0.999009
8	Violate Ethics	Comply Ethics	-2.761855	29	two-sided	0.009869	[-2.51, -0.37]	0.778646	4.554	0.984607
9	Comply Ethics	Lose Money	-0.091970	29	two-sided	0.927354	[-0.7, 0.64]	0.024552	0.195	0.051941
10	Comply Ethics	Gain Money	-0.665850	29	two-sided	0.510769	[-1.11, 0.57]	0.199690	0.238	0.184754
11	Comply Ethics	Violate Ethics	2.761855	29	two-sided	0.009869	[0.37, 2.51]	0.778646	4.554	0.984607

	p_id	variable	value
0	1	utilitarian_rank	1
34	5	deontology_rank	2
68	9	virtue_rank	2
102	13	ethics_care_rank	4
136	17	contract_theory_rank	2

	first_col	second_col	T	dof	alternative	p-val	CI95%	cohen-d	BF10	power
0	utilitarian_rank	deontology_rank	0.371023	29	two-sided	0.713317	[-0.9, 1.3]	0.108797	0.207	0.088846
1	utilitarian_rank	virtue_rank	0.441170	29	two-sided	0.662362	[-0.73, 1.13]	0.136783	0.213	0.111985
2	utilitarian_rank	ethics_care_rank	-2.953240	29	two-sided	0.006178	[-1.86, -0.34]	0.786337	6.787	0.986112
3	utilitarian_rank	contract_theory_rank	-1.966909	29	two-sided	0.058827	[-1.7, 0.03]	0.600311	1.052	0.888275
4	deontology_rank	utilitarian_rank	-0.371023	29	two-sided	0.713317	[-1.3, 0.9]	0.108797	0.207	0.088846
5	deontology_rank	virtue_rank	0.000000	29	two-sided	1.000000	[-0.93, 0.93]	0.000000	0.194	0.050000
6	deontology_rank	ethics_care_rank	-2.515813	29	two-sided	0.017667	[-2.36, -0.24]	0.773335	2.793	0.983487
7	deontology_rank	contract_theory_rank	-2.265691	29	two-sided	0.031113	[-1.97, -0.1]	0.617979	1.753	0.905038
8	virtue_rank	utilitarian_rank	-0.441170	29	two-sided	0.662362	[-1.13, 0.73]	0.136783	0.213	0.111985
9	virtue_rank	deontology_rank	0.000000	29	two-sided	1.000000	[-0.93, 0.93]	0.000000	0.194	0.050000
10	virtue_rank	ethics_care_rank	-3.555602	29	two-sided	0.001317	[-2.05, -0.55]	1.032760	25.998	0.999769
11	virtue_rank	contract_theory_rank	-3.127624	29	two-sided	0.003989	[-1.71, -0.36]	0.828765	9.886	0.992336
12	ethics_care_rank	utilitarian_rank	2.953240	29	two-sided	0.006178	[0.34, 1.86]	0.786337	6.787	0.986112
13	ethics_care_rank	deontology_rank	2.515813	29	two-sided	0.017667	[0.24, 2.36]	0.773335	2.793	0.983487
14	ethics_care_rank	virtue_rank	3.555602	29	two-sided	0.001317	[0.55, 2.05]	1.032760	25.998	0.999769
15	ethics_care_rank	contract_theory_rank	0.954773	29	two-sided	0.347584	[-0.3, 0.84]	0.227533	0.295	0.225973
16	contract_theory_rank	utilitarian_rank	1.966909	29	two-sided	0.058827	[-0.03, 1.7]	0.600311	1.052	0.888275
17	contract_theory_rank	deontology_rank	2.265691	29	two-sided	0.031113	[0.1, 1.97]	0.617979	1.753	0.905038
18	contract_theory_rank	virtue_rank	3.127624	29	two-sided	0.003989	[0.36, 1.71]	0.828765	9.886	0.992336
19	contract_theory_rank	ethics_care_rank	-0.954773	29	two-sided	0.347584	[-0.84, 0.3]	0.227533	0.295	0.225973

	utilitarian_rank	deontology_rank	virtue_rank	ethics_care_rank	contract_theory_rank
count	30.000000	30.000000	30.000000	30.000000	30.000000
mean	2.800000	2.600000	2.600000	3.900000	3.633333
std	1.584407	2.061135	1.328728	1.184672	1.159171
min	1.000000	1.000000	1.000000	2.000000	1.000000
25%	1.250000	1.000000	2.000000	3.000000	3.000000
50%	2.000000	2.000000	2.000000	4.000000	3.500000
75%	4.000000	3.000000	4.000000	5.000000	4.750000
max	6.000000	10.000000	5.000000	6.000000	6.000000

	util_rank_cat	deont_rank_cat	virtue_rank_cat	care_rank_cat	contract_rank_cat	variable	value
0	utilitarian_rank_high	deontology_rank_low	virtue_rank_high	ethics_care_rank_low	contract_theory_rank_low	loss_money_h	2.510000
1	utilitarian_rank_high	deontology_rank_low	virtue_rank_high	ethics_care_rank_low	contract_theory_rank_low	loss_money_h	0.940000
2	utilitarian_rank_high	deontology_rank_low	virtue_rank_low	ethics_care_rank_high	contract_theory_rank_low	loss_money_h	1.220000
3	utilitarian_rank_high	deontology_rank_high	virtue_rank_low	ethics_care_rank_low	contract_theory_rank_low	loss_money_h	1.600000
4	utilitarian_rank_low	deontology_rank_high	virtue_rank_low	ethics_care_rank_low	contract_theory_rank_high	loss_money_h	3.450000
...	...	...	...	...	...	...	...
135	utilitarian_rank_low	deontology_rank_high	virtue_rank_high	ethics_care_rank_low	contract_theory_rank_low	gme_y_int	-1.492078
136	utilitarian_rank_high	deontology_rank_high	virtue_rank_low	ethics_care_rank_low	contract_theory_rank_low	gme_y_int	-1.525155
137	utilitarian_rank_low	deontology_rank_high	virtue_rank_high	ethics_care_rank_low	contract_theory_rank_low	gme_y_int	-2.420982
138	utilitarian_rank_low	deontology_rank_high	virtue_rank_high	ethics_care_rank_low	contract_theory_rank_low	gme_y_int	-2.420982
139	utilitarian_rank_low	deontology_rank_high	virtue_rank_high	ethics_care_rank_low	contract_theory_rank_low	gme_y_int	-1.718643

	Utilitarianism	Deontology	Virtue Theory	Ethics of Care	Contract Theory	Violate Ethics	Comply Ethics
0	1	5	2	4	3	0.756122	1.953028
9	2	1	3	5	4	1.238374	2.266958
18	1	4	5	2	3	2.810005	0.198851
27	3	1	2	5	4	1.238374	-1.714798

	F Value	Num DF	Den DF	Pr > F
variable	4.1666	4.0000	116.0000	0.0034

	p_id	variable	value
0	1	0.9	1.0
17	18	0.9	1.0
34	5	0.8	0.4
51	22	0.8	1.0
68	9	0.2	0.3
85	26	0.2	0.2
102	13	0.1	0.0
119	30	0.1	0.0

	p_id	variable	value
0	1	0.9	1.0
17	18	0.9	1.0
34	5	0.8	0.4
51	22	0.8	1.0
68	9	0.2	0.3
85	26	0.2	0.2
102	13	0.1	0.0
119	30	0.1	0.0